search

xdit-project / xDiT

xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) on multi-GPU Clusters
Apache License 2.0
550 stars 47 forks source link

num_frames is not defined for latte #283

Closed littletomatodonkey closed 1 week ago

littletomatodonkey commented 1 week ago

Hi, when i use xDiT for latte inference, it told me the following error. I checked the diffusers and num_frames has never been a parameter for LattePipeline, so how can i avoid the problem? Thanks!

    main()
  File "/mnt/bn/multimodel/code/t2v/xDiT/examples/latte_example.py", line 40, in main
    output = pipe(
  File "/opt/tiger/miniconda3/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File "/mnt/bn/multimodel/code/t2v/xDiT/xfuser/model_executor/pipelines/base_pipeline.py", line 133, in data_parallel_fn
    return func(self, *args, **kwargs)
  File "/mnt/bn/multimodel/code/t2v/xDiT/xfuser/model_executor/pipelines/base_pipeline.py", line 147, in check_naive_forward_fn
    return self.module(*args, **kwargs)
  File "/opt/tiger/miniconda3/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
TypeError: LattePipeline.__call__() got an unexpected keyword argument 'num_frames'

My diffusers version 0.30.3, the xdit is main branch and the infer script is as follows.

torchrun --nproc_per_node=1 --master_port 12345 latte_example.py  --model="/mnt/bn/multimodel/models/t2v/maxin-cn--Latte-1/" --warmup_steps=5
feifeibear commented 1 week ago

We developped Latte using an outdated version of diffusers and plan to address this issue as soon as possible. Your assistance in resolving this matter would be greatly appreciated.